The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 38, Part II
COMPARISON AND UNCERTAINTY ANALYSIS IN REMOTE SENSING BASED
PRODUCTION EFFICIENCY MODELS
Rui Liu a, Jiu-lin Sun a, * , Juan-le Wang a, Min Liu b, Xiao-lei Li c, Fei Yang a
a
State Key Laboratory of Resources and Environmental Information Systems, Institute of Geographical Sciences and
Natural Resources Research, Chinese Academy of Sciences, No. 11A, Datun Road, Chaoyang District, Beijing 100101,
PR China - (liur, sunjl, wangjl, yangfei)@lreis.ac.cn
b
Key Laboratory of Ecosystem Network Observation and Modeling, Institute of Geographical Sciences and Natural
Resources Research, Chinese Academy of Sciences, No. 11A, Datun Road, Chaoyang District, Beijing 100101, PR
China - liusimin122@163.com
c
School of Civil Engineering and Architecture, Chongqing University of Science and Technology, Chongqing 400042,
PR China - li.xiaolei8@gmail.com
KEY WORDS: NPP, Model, PEM, Remote Sensing, Accuracy, Uncertainty, Comparsion
ABSTRACT:
The remote sensing based Production Efficiency Models (PEMs), springs from the concept of “Light Use Efficiency” and has been
applied more and more in estimating terrestrial Net Primary Productivity (NPP) regionally and globally. However, global NPP
estimats vary greatly among different models in different data sources and handling methods. Because direct observation or
measurement of NPP is unavailable at global scale, the precision and reliability of the models cannot be guaranteed. Though, there
are ways to improve the accuracy of the models from input parameters. In this study, five remote sensing based PEMs have been
compared: CASA, GLO-PEM, TURC, SDBM and VPM. We divided input parameters into three categories, and analyzed the
uncertainty of (1) vegetation distribution, (2) fraction of photosynthetically active radiation absorbed by the canopy (fPAR), (3) light
use efficiency (ε), and (4) spatial interpolation of meteorology measurements. Ground measurements of Hulunbeier typical grassland
and meteorology measurements were introduced for accuracy evaluation. Results show that a real-time, more accurate vegetation
distribution could significantly affect the accuracy of the models, since it’s applied directly or indirectly in all models and affects
other parameters simultaneously. Higher spatial and spectral resolution remote sensing data may reduce uncertainty of fPAR up to
51.3%, which is essential to improve model accuracy. We also figured out a vegetation distribution based on Maximum value of
light use efficiency (ε*) and ANUSPLIN method for spatial interpolation of meteorology measurement is also an effective way to
improve the accuracy of remote sensing based PEMs.
1. INTRODUCTION
Terrestrial net primary productivity (NPP), defined as the rate
of atmosphere carbon uptake by vegetation through the process
of net photosynthesis minus dark respiration(Ruimy et al. 1994),
is the central-related variable summarizing the interface
between plant and other processes(Field et al. 1995). NPP is
sensitive to the environmental factors and is highly various in
space and time. Thus, estimating NPP more precisely is a key to
understanding the terrestrial carbon cycle.
There are two methods available to estimate terrestrial NPP: (1)
extrapolating field measurement for local NPP to the biosphere
through a vegetation map; (2) modeling plant productivity at the
biosphere level(Ruimy et al. 1994). Since direct observation or
measurement of NPP is unavailable on a global scale, the
modeling method has been widely accepted. There are three
main types of productivity model: (1) statistical model:
estimating NPP by meteorology measurement and experimental
parameters, regardless of physiological and ecological
characteristics of vegetation, such as: Miami(Lieth 1972),
Thornthwaite(Lieth et al. 1972), Chikugo(Zenbei UCHIJIMA et
al. 1985) and Zhou Guang-sheng (Zhou et al. 1995); (2)
process model: based on plant physiological ecology principles,
estimating NPP by simulating process of photosynthesis. This
model has been widely used in local areas, such as:
CENTURY(Parton et al. 1993), CARAIB (Warnant et al. 1994,
Nemry et al. 1996), KGBM (Kergoat 1998), SILVAN (Kaduk
et al. 1996), CEVSA(Cao et al. 1998), TEM(McGuire et al.
1995) and BIOME-BGC (Running et al. 1993). (3) production
efficiency models (PEMs). The “Light Use Efficiency (ε)”
concept (Monteith 1972) has been adopted to decompose into
independent parameters such as incoming solar radiation,
radiation absorption, and conversion efficiency. The main
PEMs include CASA(Potter et al. 1993, Field et al. 1995),
GLO-PEM(Prince 1991, Prince et al. 1995), SDBM(Knorr et al.
1995), VPM(Xiao et al. 2004a, Xiao et al. 2004b) and
TURC(Ruimy et al. 1996).
Along with the increasing availability of remote sensing
measurement, (1) most parameters can be obtained by remote
sensing data, and (2)with easy access to regional data, reducing
errors caused by interpolation is possible, thus the remote
sensing based PEMs has been applied more and more to
estimatting terrestrial NPP.
However, it is very difficult to evaluate the accuracy of the
models for two reasons: (1) acquisition of direct observation is
* Corresponding author. Tel.: +86-10-64889266; +86-10-64889045; Fax.: +86-10-64889062
E-mail addresses: sunjl@lreis.ac.cn (Jiu-lin Sun) liur@lreis.ac.cn (Rui Liu)
185
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 38, Part II
unavailable on a regional or global scale, and (2) different
models come from different data sources and handling methods,
impossible to determine which one is more accurate.
simulation process, remote sensing data mainly provide the
following three types of information:
Model
Under the scientific sponsorship of the IGBP, such a model
intercomparison has been carried out at the Potsdam Institute
for Climate Impact Research(Cramer et al. 1999, Ruimy et al.
1999). Result shows that global NPP estimates (Fig.1) vary
greatly between different models. However, since no field data
are available to validate the models, we have no way to
determine which result is closer to “true value”.
Influenced by:
NPP=f (ε*, RS, fPAR, T, EET, PET )
CASA
GLO-PEM
NPP=f (ε*, RS, fPAR, T, VPD, SW, RA )
TURC
NPP=f (ε*, RS, fPAR, RA, T )
SDBM
NPP=f (ε*, RS, fPAR, CO2 )
NPP=f (ε*, RS, fPAR, T, W, PL )
VPM
RS: Solar radiation
RA: Plant autotrophic respiration
PET: Potential evapotranspiration
EET: Estimated evapotranspiration
PL: Leaf phenology
T: Temperature
W: Water capacity
Table 1. Parameters of PEMs
Model
CASA
Vegetation
Satellite
DistrifPAR
bution
×
GLO-PEM
TURC
RS,T,EET,PET
×
×
Plant
Physioecology
Other
satellite data
ε*
*
ε ,RA
×
SDBM
VPM
×
Meteorological
measurements
RS,T,VPD,SW
*
×
RS,T
ε ,RA
×
RS,CO2
ε*
×
RS,T
ε*,PL
W
Figure 1. Comparison of global NPP estimations of PEMs
Table 2. Parameters classification of PEMs
In this study, we focus on the methods of improving the
accuracy of remote sensing based on PEMs rather than
determining the better one. The input parameters of PEMs are
divided into three categories according to acquisition source
and the analysis of of each category’s influence is performed.
We compare the 5 models by taking different data sources as
input for each parameter and
combining with ground
measurement in Hulunbeier and Tibet in order to determine the
uncertainty of parameter. The purpose of this study is, first, to
identify the most influential input parameters in NPP estimation
among the 5 models, and second, to seek access to improvement
in model accuracy.
1. Vegetation Distribution Information: According to
different spectral characteristics and temporal variation, the
accurate and real-time vegetation distribution on the earth
can be obtained through appropriate classification
algorithm.
2. Vegetation Index: Spectral reflectance of vegetation is
influenced by vegetation type, species composition,
vegetation cover, chlorophyll content, plant water and so
on. Vegetation index is a comprehensive performance of
the spectral reflectance, which bears a strong relationship
with NPP estimates.
3. Vegetation growth environment information: The
environmental factors can be obtained by means of remote
sensing in recent years, such as temperature, precipitation,
soil moisture and other relevant information, though
further study is to be made on the applicability and
accuracy .
2. PARAMETERS ANALYSIS IN PEMs
PEMs develops from the concept of light use efficiency: NPP
has a strong linear relationship in ideal environment with light
use efficiency (ε) and absorbed photosynthetically active
radiation (APAR): NPP=ε ∙APAR.
GLO-PEM is unique among the 5 models because all variables
about climate and vegetation distribution are derived from
remote sensing data.
APAR is calculated from global solar radiation and fraction of
photosynthetically active radiation absorbed by the canopy
(fPAR) which can be obtained from remote sensing data, and ε
is regarded as a conversion scale of APAR to NPP, as a result of
the interaction of environmental constraints and based on
“Maximum value of light use efficiency (ε*)”. These models
include various parameters (Tab.1) and focus on different levels
of mechanism with inhibition process taken into account. The
parameters can be divided into three categories (Tab.2)
according to acquisition source.
2.2 Meteorology Measurement
The process of plant growth appears responsive to
environmental conditions. Therefore, the formation of
vegetation NPP depends on the regional light, heat, water
conditions and so on, as well as the biome production
capacity(Zhou et al. 1995). Climate factors’ control over
vegetation productivity is not only present in the vegetation
diversity, but also in photosynthesis inhibition. The
meteorology measurements in the models such as radiation,
temperature, precipitation are all obtained from meteorology
stations except GLO-PEM.
2.1 Remote Sensing Data
PEMs uses remote sensing data to acquire the land surface
condition, especially vegetation type and fPAR. In the
186
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 38, Part II
2.3 Plant Physiological Data
3.2 Remote Sensing Based fPAR
The plant physioecology mainly concerns how and to what
extent the plant growth responds to the environmental factors
such as: increasing CO2 concentration, ultraviolet radiation
enhancement, temperature change, sunlight irradiation and the
enlargement of salty habitats. All of these factors are closely
associated with the process of global climate change.
fPAR, a significant parameter for calculating APAR, truly
reflects the status of vegetation canopy’s absorption of
photosynthetically active radiation, and has a direct impact on
the uncertainty of PEMs. Remote sensing provides a means to
estimating fPAR globally. Most models (CASA, GLO-PEM,
SDBM and TURC) utilizes Normalized Difference Vegetation
Index (NDVI) to obtain fPAR (apply different algorithms), while
VPM uses Enhanced Vegetation Index (EVI) (Tab.3).
ε is a fundamental element in PEMs. It was used for the
conversion of APAR to biomass, and affected by many
environmental factors. Each model illustrates its own approach
of simulating the process in which environmental factors
influence mode.
CASA
fPAR=min{(SR−SRmin)/(SRmax−SRmin), 0.95}
GLO-PEM fPAR=(SR−SRmin)(fPARmax−fPARmin)/(SRmax−SRmin)
TURC
fPAR= −0.1914+2.186∙NDVI
SDBM
fPAR= −0.025+1.25∙NDVI
VPM
fPAR=1∙EVI
3. UNCERTAINTY OF INPUT PARAMETERS
SR=(1+NDVI)/(1−NDVI)
Obviously enough, parameters are highly similar in these
models. The main differences lie in (1) the way of obtaining and
applying vegetation distribution, (2) the way of obtaining fPAR
and ε and (3) the use of meteorology factors. Thus, it is
important for improving model accuracy to analyze uncertainty
of each parameter.
Table 3. fPAR estimation in PEMs
However, some researches indicate that there are limitations in
application of NDVI, such as (1) tending to be saturated in wellvegetation cover area(Wang et al. 2003), and (2)which is
sensitive to the soil structure in the low vegetation cover
area(Huete et al. 1994). One possible solution is using EVI
which introduced the blue-ray band aiming to reduce
atmosphere effect(Huete et al. 1994, Huete et al. 1997).
Although EVI was used in VPM for estimating forest NPP, and
considered superior in grassland NPP estimation over
NDVI(Kawamura et al. 2005), the further application in
different environment on a global scale is still necessary.
3.1 Vegetation Distribution
Vegetation distribution is considered to be the most important
determinant of carbon storage, uptake and release from the
terrestrial biosphere, and it affects model accuracy mainly in
two ways:
3.1.1 Applying Vegetation Distribution: All remote sensing
based PEMs assumes the world is covered by vegetation. CASA
and VPM uses an actual vegetation distribution from remote
sensing including human land use. SDBM and GLO-PEM does
not use a vegetation map directly, but the parameters via remote
sensing such as temperature , vapour pressure deficit and APAR
also explain the vegetation cover change. Only TURC uses
potential vegetation regardless the human land use. A
comparison between actual vegetation data set and potential
vegetation data set shows that the human land use and
agriculture affect up to 40% of NPP estimate in temperate
mixed forests and deciduous forest on a global scale(Ruimy et
al. 1999). Regionally the simulated NPP with land use
constraint in the south portion of NSTEC was about 65% of that
without land use constraint(Gao et al. 2003). On the other hand,
a better classification accuracy has testified its improvement for
NPP estimate(Zhu et al. 2006).
It remains unsolved which fPAR is more accurate in these
models since we cannot have fPAR from ground measurement.
However, we can improve the precision by using higher spatial
resolution NDVI and EVI. All the models calculate the fPAR
with a 8km resolution NDVI derived from NOAA/AVHRR. For
the ease of comparison, we use MODIS NDVI and EVI data
(obtained at 2009-07-28 from USGS) with 1km, 500m and
250m spatial resolution, and ground measured spatial data
(obtained at 2009-08-02 by FieldSpec® HandHeld
Spectroradiometer) from Hulunbeier grassland (Fig.2). The
calculation is strictly in accordance with the MODIS algorithm.
It should be known that we are not saying the ground measured
data is “true”, but it is of great reference value. The relative
error (Fig.3) shows that the better spatial resolution brings the
higher precision. In some extra point, relative error from the
1km resolution can reach up to 51.3%.
3.1.2 Determination of other parameters: Vegetation
distribution is applied directly or as an intermediate variable to
determine the precision of other parameters such as ε*, RA, PL
and EET. In most models, these parameters are assumed
constant or determined by the vegetation maps. However, these
plant physiological-related parameters apparently depend on the
vegetation type with the classification accuracy taken into
account.
It is impossible for us to know exactly how much vegetation
distribution is affected, but we definitely know that a real-time,
more accurate vegetation distribution can significantly affect the
accuracy of the models.
187
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 38, Part II
3. More accurate remote sensing data are uesd for
retrieval and plant ecology data have been counted for a
more accurate ε*.
3.4 Spatial Interpolation of Meteorology Measurements
Based on different models, climate factors affect the NPP
estimation working as the inhibition of ε* (Tab.4). However,
they are hardly obtained directly through remote sensing data
with high accuracy.
In most cases, the regional and global meteorology distribution
are based on station measurement and spatialized by means of
interpolation. Therefore, interpolation precision is important in
improving accuracy of NPP estimates(Price et al. 2000, Wong
et al. 2003). The precision of meteorology interpolation is
mainly influenced by: (1) site’s latitude and longitude, (2) site’s
elevation and (3) regional terrain.
Figure 2. Comparison of NDVI and EVI
Model
CASA
GLO-PEM a
TURC
SDBM
VPM
a
Influenced by
RS, T, EET, PET
RS, T, VPD, SW
RS, T, RA
RS, CO2
RS, T, W
All factors in GLO-PEM are obtained from remote sensing data
Table 4. Climate factors in PEMs
A comparisive study of several interpolation method shows that
the multiple regression equation had a better performance by
introducing elevation and location(Collins 1995). Lin’s research
demonstrated that Gradient Plus Inverse-distance-squared
(GIDS) method is better than others in reflecting temperature
change with elevation(Lin et al. 2002), but researchers also
proved that ANUSPLIN method is better than GIDS(Price et al.
2000, Feng 2004). The 88 meteorology station data in Tibet was
spatialized using ANUSPLIN method (Liu 2008) combining
with 1km×1km DEM. The result (Fig.4) proved to be more
accurate than other methods.
Figure 3. Relative error of NDVI and EVI
3.3 Estimating Maximum Value of Light Use Efficiency
The light use efficiency (ε) determines the capability that the
plants capture and transform environmental resources to dry
matter production, and fluctuate with the environmental index
such as temperature, moisture, soil, nutrition, plant ontogeny,
etc. (Prince 1991). In remote sensing based PEMs, these
affections present as constraints of Maximum Value of Light
Use Efficiency (ε*) ranging between 0 and 1. Therefore, ε*is
influential for NPP estimates.
In early researches, ε* is empirically derived as a conservative
quantity(Monteith 1972). CASA takes it as 0.389gC∙MJ−1.
However, several researches indicate that ε* varies due to
different vegetation types. And in view of its importance, there
is controversy about the value range. Ruimy believes it ranges
between 0.108 and 1.580 gC∙MJ−1 and GLO-PEM adopts it
between 0.2 and 1.2 gC∙MJ−1. In Guangdong Province of China,
the result shows that ε* range between 0.69 and 1.05 gC∙MJ−1
(Peng et al. 2000). Since ε cannot be measured directly, studies
about determination of ε generally fall into two types: (1)
simulating the plant growth process with the principle of plant
ecology, and (2) remote sensing retrieval through PEMs and
ground measured NPP. The latter one seems more feasible as
plant ecology simulation can hardly be extended to a global
scale. By means of remote sensing retrieval, there are methods
for improving the precision of PEMs:
1. ε* should not be considered as a conservative quantity,
for it shows difference between different biome.
2. A more accurate and real-time vegetation map could
be used which shows the biome distribution.
188
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 38, Part II
sensing retrieval with plant ecology. Nowadays, ε* can be
obtained precisely by measuring fluxes of CO2 over whole
canopies, and analyzing the relationship between CO2 exchange
and photon flux density. It is generally possible to extract ε*
more representatively though the uncertainty remains to be
developed.
Being as constraints of ε*, meteorology measurement and ε*
determine the plant conversion efficiency together. Spatial
interpolation method determines the accuracy of spatialized
meteorology data. The chosen method should involve all
relating factors, including location, elevation and terrain. Our
research shows that the ANUSPLIN method can improve
accuracy.
Here, we developed methods to improve parameter accuracy,
though the overall accuracy improvement for the model still
remains unquantified. Meanwhile, although we are not sure
about the interrelation response mechanism between each
parameter, it is possible to estimate NPP at a higher precision
through applying more accurate parameters.
Figure 4. Meteorology interpolation in Tibetan transact
4. CONCLUSION AND DISCUSSION
Remote sensing based PEMs takes good use of the “light use
efficiency” theory and adopts several approaches to estimate
NPP. Each approach has a theoretical basis for its own. It is not
possible to tell which model is “better” since none of them is
perfect and cannot be verified on a global scale. NPP estimate
varies greatly among models in a comparative research and we
have no way to know which result is closer to the “true value”.
However, there are ways to improve the model accuracy.
ACKNOWLEDGEMENTS
We greatly appreciate the support of Nature Science Founding
of China (40771146, 40801180) and Open Research Funding
Program of LGISEM.
REFERENCE
Vegetation distribution is the fundamental element among all
parameters and has been used directly or indirectly in all
models. Apparently, actual vegetation distribution performs
much better than potential vegetation distribution while human
land use has a great impact on the NPP estimation. Meanwhile,
vegetation distribution determines the accuracy of the
application of other parameters to a large extent. The
uncertainty of vegetation distribution is caused by: (1)
inconsistent and ambiguous vegetation types, (2) time
inconformity from classification time to current time, (3) mixed
pixel of different vegetation types and (4) inconsistent scaling.
The developing remote sensing data and techniques provide the
possibility to these uncertainties. Overall, a real-time and
accurate vegetation map is helpful in greatly improving the
accuracy of the PEMs.
Cao, M.K. and Woodward, F.I., 1998, Dynamic responses of
terrestrial ecosystem carbon cycling to global climate change.
Nature, 393, pp. 249-252.
Collins, F.C., and P.V. Bolstad. 1996. A comparison of spatial
interpolation techniques in temperature estimation. In:
Proceedings of the Third International Conference/Workshop
on Integrating GIS and Environmental Modeling, Santa Fe,
New Mexico, January 21-25, 1996. Santa Barbara, California:
National Center for Geographic Information Analysis (NCGIA).
Cramer, W., Kicklighter, D.W., Bondeau, A., Moore, B.,
Churkina, C., Nemry, B., Ruimy, A., Schloss, A.L. and
Participants Potsdam, N.P.P.M.I., 1999, Comparing global
models of terrestrial net primary productivity (NPP): overview
and key results. Global Change Biology, 5, pp. 1-15.
The vegetation index is close to photosynthesis by means of
determining the fPAR. The NOAA/AVHRR NDVI is the most
common data for NPP estimate. But the 8km or lower spatial
resolution caused large errors because of mixed pixel.
Advanced sensors with better spectral and spatial resolution can
provide more accurate fPAR. Our experiment in Hulunbeier
shows that the better resolution brings higher precision,
especially in mixed pixels which have 51.3% relative error.
Feng, X.F., 2004, Simulating Net Primary Productivity and
Evaportranspiration of Terrestrial Ecosystems in China using a
Process Model Driven by Remote Sensing. Institute of
Geographical Sciences and Natural Resources Research,
Chinese Academy of Sciences.
Field, C.B., Randerson, J.T. and Malmstrom, C.M., 1995,
Global Net Primary Production - Combining Ecology and
Remote-Sensing. Remote Sensing of Environment, 51, pp. 7488.
New vegetation index was introduced for NPP estimate.
Although EVI proves better to perform vegetation status than
NDVI, there are still questions about the interrelationship
between fPAR and EVI. It is regarded as a potential solution and
needs more research work.
Gao, Q., Li, X.B. and Yang, X.S., 2003, Response of
Vegetation and Primary Production in Northern-South Transect
of Eastern China to Global Change Under Land Use Constraint.
Acta Botanica Sinica, 45, pp. 1274-1284.
ε* varies greatly among literatures for there is no convincing
method for ground measuring or evaluation. Since ε* depends
on vegetation types, we can enhance the model accuracy by (1)
using more accurate vegetation map and (2) combining remote
189
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. 38, Part II
Potter, C.S., Randerson, J.T., Field, C.B., Matson, P.A.,
Vitousek, P.M., Mooney, H.A. and Klooster, S.A., 1993,
Terrestrial Ecosystem Production - A Process Model-Based on
Global Satellite and Surface Data. Global Biogeochemical
Cycles, 7, pp. 811-841.
Huete, A., Justice, C. and Liu, H., 1994, Development of
vegetation and soil indices for MODIS-EOS. Remote Sensing of
Environment, 49, pp. 224-234.
Huete, A., Liu, H., Batchily, K. and Van Leeuwen, W., 1997, A
comparison of vegetation indices over a global set of TM
images for EOS-MODIS. Remote Sensing of Environment, 59,
pp. 440-451.
Price, D., McKenney, D., Nalder, I., Hutchinson, M. and
Kesteven, J., 2000, A comparison of two statistical methods for
spatial interpolation of Canadian monthly mean climate data.
Agricultural and Forest Meteorology, 101, pp. 81-94.
Hutchinson, M., 2003, ANUSPLIN version 4.2 user guide.
Centre for Resource and Environmental Studies, Australian
National University, Canberra, Australia.
Prince, S., 1991, A model of regional primary production for
use with coarse resolution satellite data. International Journal
of Remote Sensing, 12, pp. 1313-1330.
Kaduk, J. and Heimann, M., 1996, A prognostic phenology
scheme for global terrestrial carbon cycle models. Climate
Research, 6, pp. 1-19.
Kawamura, K., Akiyama, T., Yokota, H., Tsutsumi, M., Yasuda,
T., Watanabe, O., Wang, G. and Wang, S., 2005, Monitoring of
forage conditions with MODIS imagery in the Xilingol steppe,
Inner Mongolia. International Journal of Remote Sensing, 26,
pp. 1423-1436.
Kergoat, L., 1998, A model for hydrological equilibrium of leaf
area index on a global scale. Journal of Hydrology, 213, pp.
268-286.
Ruimy, A., Dedieu, G. and Saugier, B., 1996, TURC: A
diagnostic model of continental gross primary productivity and
net primary productivity. Global Biogeochemical Cycles, 10, pp.
269-285.
Ruimy, A., Kergoat, L., Bondeau, A. and Intercomparison,
P.P.N.M., 1999, Comparing global models of terrestrial net
primary productivity (NPP): analysis of differences in light
absorption and light-use efficiency. Global Change Biology, 5,
pp. 56-64.
Ruimy, A., Saugier, B. and Dedieu, G., 1994, Methodology for
the estimation of terrestrial net primary production from
remotely sensed data. Journal of Geophysical ResearchAtmospheres, 99, pp. 5263-5283.
Knorr, W. and Heimann, M., 1995, Impact of Drought Stress
and Other Factors on Seasonal Land Biosphere CO2 Exchange
Studied Through An Atmospheric Tracer Transport Model.
Tellus Series B-Chemical and Physical Meteorology, 47, pp.
471-489.
Lieth, H., 1972, Modeling the primary productivity of the world.
Nature and Resources, 8, pp. 5-10.
Running, S. and Hunt, E., 1993, Generalization of a forest
ecosystem process model for other biomes, BIOME-BGC, and
an application for global-scale models. Scaling physiological
processes: leaf to globe, pp. 141-158.
Lieth, H. and Box, E., 1972, Evapotranspiration and primary
productivity; CW Thornthwaite Memorial Model. publications
in climatalogy, 25, pp. 37-46.
Wang, Z.X., Liu, C. and Alfredo, H., 2003, From AVHRRNDVI to MODIS-EVI: Advances in Vegetation Index Research.
Acta Ecological Sinica, 23, pp. 979-987.
Lin, Z.H., Mo, X.G., Li, H.X. and Li, H.B., 2002, Comparison
of Three Spatial Interpolation Methods for Climate Variables in
China. ACTA GEOGRAPHICA SINICA, 57, pp. 47-56.
Warnant, P., Francois, L., Strivay, D. and Gerard, J.C., 1994,
Caraib: A Global-Model of Terrestrial Biological Productivity.
Global Biogeochemical Cycles, 8, pp. 255-270.
Liu, M., 2008, Study on Estimation and Uncertainty of
Terrestrial Ecosystem Productivity Based on RS and GIS.
Nanjing Normal University.
Xiao, X., Hollinger, D., Aber, J., Goltz, M., Davidson, E.,
Zhang, Q. and Moore, B., 2004a, Satellite-based modeling of
gross primary production in an evergreen needleleaf forest.
Remote Sensing of Environment, 89, pp. 519-534.
McGuire, A.D., Melillo, J.M., Kicklighter, D.W. and Joyce,
L.A., 1995, Equilibrium responses of soil carbon to climate
change: Empirical and process-based estimates. Journal of
Biogeography, 22, pp. 785-796.
Xiao, X., Zhang, Q., Braswell, B., Urbanski, S., Boles, S.,
Wofsy, S., Moore, B. and Ojima, D., 2004b, Modeling gross
primary production of temperate deciduous broadleaf forest
using satellite images and climate data. Remote Sensing of
Environment, 91, pp. 256-270.
Monteith, J.L., 1972, Solar-Radiation and Productivity in
Tropical Ecosystems. Journal of Applied Ecology, 9, pp. 747766.
Zenbei UCHIJIMA and SEINO, H., 1985, Agroclimatic
Evaluation of Net Primary Productivity of Natural Vegetation
(1) Chikugo Model for Evaluating Net Primary productivity. J.
Agr.Met, 40, pp. 343-352.
Parton, W.J., Scurlock, J.M.O., Ojima, D.S., Gilmanov, T.G.,
Scholes, R.J., Schimel, D.S., Kirchner, T., Menaut, J.C.,
Seastedt, T., Moya, E.G., Kamnalrut, A. and Kinyamario, J.I.,
1993, Observations and Modeling of Biomass and Soil OrganicMatter Dynamics for The Grassland Biome Worldwide. Global
Biogeochemical Cycles, 7, pp. 785-809.
Zhou, G.S. and Zhang, X.S., 1995, A Natural Vegetation NPP
Model. Acta Phytoecologica Sinica, pp. 193-200.
Zhu, W.Q., Pan, Y.Z., He, H., Yu, D.Y. and Hu, H.B., 2006,
Maximum Value of Light Use Efficiency Simulation of China's
Typical Vegetation. CHINESE SCIENCE BULLETIN, 51, pp.
700-706.
Peng, S.L., Guo, Z.H. and wang, B.S., 2000, Use of GIS and RS
to Estimate the Light Utilization Efficiency of the Vegetation in
Guangdong, China. Acta Ecological Sinica, 20, pp. 903-909.
190